One AI Engineer Intern candidate singled out the Web App portion of an upcoming final.

Coverage: One Klaviyo AI Engineer Intern update from November 2025, followed by 14 separate Klaviyo SWE/SDE and full-stack internship reports published in 2021–2026.

AI Engineer Intern: an educated guess

I'd expect Python backend coding and debugging, AI application work, and a deep dive into a project you've built.

The intern candiate report below mentions a Web App portion. Klaviyo's Summer 2027 AI Engineer Intern job description also calls for backend services, Python, FastAPI/Django, agents, retrieval-augmented generation (RAG), and model evaluation. That's a newer hiring cycle, so we're using it to judge the role's direction.

A practice example we'd use: (not a reported question:) Start with an existing Python service. Add an endpoint that calls an LLM to classify customer messages. Check the response format, save the result, handle failures, and add tests. Then explain how you'd measure classification quality.

We'd put backend/API work, AI projects, and evaluation first. Model training and reinforcement learning appear as nice-to-haves in that job description, so we'd spend less time on heavy math and training a model from scratch.

AI Engineer Intern — the confirmed update

Report date: November 17, 2025 · Status: Final round coming up

The candidate explicitly named the AI Engineer Intern role. They said their final round was approaching and that they were most concerned about the Web App portion. They asked another candidate what to expect.

They did not share a coding prompt, framework, interview date, or later result. This is a final-round update, not a completed technical debrief.

The Stripe webhook task below comes from a Klaviyo SWE Intern report. It has not been verified for AI Engineer Intern.

What came up in the SWE Intern reports

The 14 numbered accounts below cover Klaviyo software engineering internships. Their tasks include web applications, debugging, database work, and recursion. In report 01, Stripe is the webhook technology named in a Klaviyo coding exercise; the candidate was interviewing with Klaviyo.

A candidate can appear in more than one row. The AI Engineer Intern update is separate from these counts.

Online assessments

OA 01: Two 2024 descriptions

1. Report 03, November 2024: The candidate described an online OA followed by a three-hour virtual final. The final included two hours of Python/SQL web-development interviews and a behavioral session. They did not share the OA questions.
2. Report 04, October 2024: The candidate described a LeetCode-like OA, followed by Python API-design work in the final round. They accepted an offer.

Early technical screens

Screen 01: Debugging a project with two engineers

Interview date: September 20, 2021 · Role: SDE intern

1. The candidate worked on an existing project, fixing bugs and adding features.
2. Two engineers watched the session.
3. The candidate could ask questions when something was unclear.

They described it as a collaborative work simulation and had a positive impression of the format. This is the first stage of candidate report 14 below.

SWE Intern questions and tasks

Web applications and databases

Weather API and frontend work — report 02. The intern described building a weather app that uses an API and writing frontend code during a three-hour superday. The weather API practice project covers that task shape. The report did not name a language, framework, or API provider.

1. Weather app, 2023 — report 08. The task had four parts, each adding a small piece of functionality. The candidate said finishing every part was not required and that communication mattered. They did not preserve the individual requirements. The related weather-app exercise focuses on saving email and city pairs.
2. Flask application, 2022 — report 10. The candidate worked with Flask, the Python Requests library, REST APIs, SQLite, and Pony ORM, a library for working with database records through Python objects. They completed two of four tasks. The interviewer said most candidates did not have time for the other two.
3. Flask and Requests, 2022 — report 09. These appeared in both the first technical interview and one final-round session. The candidate did not describe the exact features or API behavior.

Debugging and code review

1. Stripe webhook task at Klaviyo — SWE Intern report 01. The candidate named creating and fixing Stripe webhooks, without preserving the original code or defects. Our related webhook-handler exercise supplies a small application to repair.
2. Python debugging — report 05. Correct bugs, consider whether SQL injection is possible, then explain what else could improve once the bugs are fixed.
3. API and HTTP calls — report 06. Work with the interviewer to find bugs and improve a real-world code sample, including API calls, HTTP client/server calls, and basic SQL.
4. Pair programming — report 07. The candidate described system-design discussion and debugging, followed by a question about optimizing the code. They did not identify its application domain.
5. Two API calls and a database update — report 13. The candidate also faced a follow-up about handling millions of requests. The post does not preserve the API endpoints or data model.

Recursion and follow-ups

1. JSON to key-value pairs — report 08. The input was JSON, and the task involved recursively converting it into key-value pairs. The candidate's first approach was not optimal. The interviewer asked for an improvement and helped them reach a better solution. A follow-up flipped the input and output. The exact data format and reverse-conversion rules were not recorded.

2. A medium recursion problem — report 09. The candidate could not identify a matching LeetCode problem. No fuller prompt was given.

3. Recursion, test cases, and Python — report 10. The candidate solved the problem quickly, then wrote test cases and discussed their understanding of Python. They initially described the problem as medium to somewhat difficult, then clarified in a reply that it was a fairly standard medium. They also said it was presented directly as a problem, without a larger background story.

4. Nested JSON — report 11. A separate 2022 Reddit candidate described breaking down nested JSON with recursion. They did not preserve the required output format, so this account does not identify an exact match to report 08's key-value-pair task.

Past projects and architecture

1. Report 08: A detailed discussion of past projects, why the candidate wanted to join Klaviyo, and how they learned something quickly. The candidate felt the interviewers could follow technical details closely.
2. Report 09: A manager conversation about previous projects. The post did not preserve individual questions.
3. Report 10: Draw a multi-tier application's local architecture, turn it into an AWS cloud design, and discuss improvements using Redis, S3, and an API gateway. The candidate later said their resume listed AWS and GCP architecture certifications, which they thought might explain this line of questioning. Two behavioral questions followed, but their wording was not shared.

14 SWE Intern candidate timelines, loops, and screens

01: Klaviyo SWE Intern — a Stripe webhook task in a 3.5-hour interview

Report date: February 12, 2026 · Role: Software engineer intern · Result: No offer

1. Three consecutive interviews: two technical and one behavioral, totaling about 3.5 hours.
2. Klaviyo's reported coding task involved creating and fixing Stripe webhooks.
3. The candidate found the technical sessions challenging but said the interviewers were interactive and did not expect every answer to be correct.

02: A weather API and frontend task

Report date: March 17–18, 2026 · Role: Full-stack software engineering intern

An intern replying to a final-round discussion said they had interviewed the previous September. They described a three-hour superday that included implementing a weather app using an API and writing frontend code. When asked, they confirmed it was an internship interview.

03: An OA followed by a three-hour virtual final

Report date: November 21, 2024 · Role: Software engineer intern · Result: No offer

1. An online OA came first.
2. The candidate described two hours of technical interviews involving web development with Python and SQL, plus one behavioral session.
3. One reported question asked what they were looking for in an internship.

04: Python API design after an algorithm-style OA

Report date: October 22, 2024 · Role: Software engineer intern · Location: Boston · Result: Accepted offer

The candidate described the OA as LeetCode-like. The final round focused more on designing APIs in Python, which they compared with system design. A behavioral question asked about a time they had collaborated. The post did not preserve an exact API contract or a standalone algorithm prompt.

05: Python bugs, SQL injection, and the next improvement

Report date: October 7, 2024 · Role: Software engineer intern · Result: No offer

HR contacted the candidate. Their first interview involved debugging a Python program, including details such as whether SQL injection was possible. After the bugs were fixed, the interviewer asked what further improvements they could make. The report did not identify the program's business domain.

06: A career-fair candidate's first collaborative interview

Report date: October 30, 2023 · Role: Software engineer intern · Status: One interview completed at the time of the post

The candidate got the interview through a university career fair. They worked with the interviewer to find bugs, issues, and improvements in a real-world code sample. Their reported topics included API calls, HTTP client/server calls, and basic SQL. They described the session as very collaborative; no later outcome was given.

07: Pair programming before a planned three-stage final

Report date: October 24, 2023 · Role: Software engineer intern

The candidate described a phone screen and pair programming, with a three-stage virtual final of about three hours still ahead. The pair session involved system-design discussion and debugging. One follow-up asked how to optimize the code. The post does not describe completing the final round or explain an eventual result.

08: A 2023 offer after a train task, weather app, and JSON recursion

Report date: October 19, 2023, with updates through October 25 · Role: SWE intern · Result: Verbal offer

Recruiter and first technical round

1. September 7: Applied.
2. September 18: Received an invitation to a recruiter phone screen.
3. September 19: Took the recruiter screen and scheduled the first technical interview.
4. September 28: Completed the first collaborative coding round. The candidate identified it as the train problem discussed in other posts, without repeating its requirements.

Final round and result

1. October 2: Received the final-round invitation.
2. October 18: Completed the final round. Their description included a four-part weather app, JSON recursion, and a manager conversation about projects and behavioral questions.
3. In the recursion session, the interviewer challenged the first approach, helped with an improvement, and asked about reversing the conversion.
4. October 24: Received a verbal offer by phone.

The candidate enjoyed the interviews and felt the engineers had broad technical knowledge. They corrected an early timeline summary to say the whole process took less than two months.

09: A 2022 offer after three 30-minute final interviews

Report date: November 15, 2022 · Role: Software engineering intern · Result: Offer call and letter

Recruiter and first technical round

1. October 17: A recruiter reached out.
2. October 21: The first interview involved Flask and Requests.

Final round and result

1. November 9–10: Three final interviews, each about 30 minutes.
2. One session involved Flask and Requests again.
3. Another was a medium recursion problem. The candidate did not know a matching LeetCode number.
4. The manager session covered previous projects.
5. November 15: Received the offer call and letter.

10: Two of four app tasks completed, then a rejection

Report date: November 4, 2022 · Role: SDE intern · Result: Rejected

Early rounds

1. Applied through campus recruiting.
2. October 8: A behavioral interview that the candidate found straightforward.
3. October 15: A small program using Flask, Requests, REST APIs, and SQLite. Time was tight, and the program also needed improvements.

October 21 final round

1. An application task using those tools plus Pony ORM. The candidate finished two of four parts; the interviewer said most people did not have time for the remaining two.
2. A recursion problem, followed by test cases and a discussion about Python.
3. A conversation about previous experience and application architecture: draw the local system, move it to AWS, then discuss Redis, S3, and an API gateway. Two behavioral questions followed.

The candidate felt the interviews had gone reasonably well but was rejected about a week later. They wondered whether headcount played a part; the post does not establish the reason. Their later replies also clarify that the architecture questions may have been prompted by the cloud certifications on their resume.

11: A 2022 campus-fair process with three technical sessions

Report period: November 2022 · Role: SWE intern · Result: Rejected

The candidate applied after a campus job fair on September 28, received an invitation on September 29, and interviewed on October 1 and 12. An automated rejection arrived October 14.

Three technical sessions took about four hours. The first two involved learning JSON and what the author called React on the spot. Their Python/React wording was unclear, and the individual tasks were not preserved. The third session involved breaking down nested JSON with recursion.

They appreciated the quick process but were disappointed by an automated rejection after that much interview time.

12: A small-ticket coding session and an offer

Report date: January 19, 2022 · Role: Software engineering intern · Result: Accepted offer

1. An introductory call led to a technical interview.
2. In that interview, the candidate worked through a small ticket with another engineer.
3. The final consisted of two technical interviews and one conversation with the head engineer.
4. A reported question asked about their process for debugging broken code.

13: Two API calls, database updates, and a scaling follow-up

Report date: November 2, 2021 · Role: Software engineer intern · Result: No offer

The candidate applied through their university. The technical task involved writing two API calls and updating data in a database. They thought the task looked simple but found the interviewer's expectations unclear at first. A follow-up asked how the application would handle millions of requests. Their overall reaction was negative despite rating the interview easy.

14: A 2021 career-fair application that moved quickly

Report date: October 21, 2021 · Role: SDE intern · Result: Offer call

Application and first screen

1. September 13: Spoke with a Klaviyo engineer at a school career fair and applied online that day.
2. September 16: Received the first interview invitation.
3. September 20: Completed the collaborative work simulation described above: debug a project and add features while two engineers observed.

Final round and result

1. September 21: Received the final-round invitation.
2. September 27: Completed algorithm, collaborative coding, and behavioral sessions.
3. The candidate placed the algorithm question between easy and medium. They did not share the actual prompt.
4. September 28: Received an offer call.

They described the process as efficient and the interview experience as very positive.

Patterns across the SWE Intern reports

1. Three recent accounts describe finals lasting three to 3.5 hours. Reports 01, 02, and 03 describe those durations. Report 09's older final was three 30-minute interviews across two dates.
2. The web tasks had different scopes. Report 01 names a Stripe webhook exercise at Klaviyo, report 02 names a weather API and frontend work, and report 04 describes Python API design. Report 08's four-part weather task does not preserve enough detail to identify the same application as report 02.
3. Partial completion did not establish a pass rule. Report 08 emphasized communication, report 10 records two of four app tasks completed, and report 01 says perfect answers were not expected. Their outcomes differ, and none gives a scoring cutoff.